Skill 03
IT security & reverse engineering
Security that does not rely on nobody ever pressing the wrong button.
I prefer to solve security problems in a way that makes them impossible in the first place: an AI agent that only knows the tools of its task cannot do anything else, even when it reads a crafted email. A server configured deny-by-default does not have to fend off every new hole individually.
Where the job is to check rather than build, I use tools that prove instead of guess – symbolic execution to show that two functions do the same thing, or static analysis of installers without ever running them.
What I do
- Reverse engineering and analysis of Windows binaries
- Checking installers and software packages before rollout
- Hardening Linux servers and container environments
- Security concepts for AI agents and LLM integrations
- Log analysis and alerting on attacks
Tech stack
| Layer | Technology | Why |
|---|---|---|
| Binary analysis | angr, VEX, Z3 | proofs instead of samples |
| Formats | PE/COFF, ELF, MSI/OLE, WiX Burn | analysis on Linux, no Windows API |
| Network | Suricata, Filebeat | detect attacks in traffic |
| Operations | OpenObserve alerts, audit logs | trace who tried what and when |
| AI security | policy engine, task scopes, fixed recipients | rule out prompt injection by construction |
Related projects
- 01pe-analyzerProves whether two binary functions do the same thingReverse engineering tool
- 02install-parserUnderstand Windows installers without running themAnalysis tool for software packaging
- 03OpenHarnessAn AI assistant that rules out prompt injection by constructionSelf-hosted AI agent
- 04HomelabProxmox, Nextcloud, Pi-hole, Homebridge and central loggingSelf-hosted infrastructure
Planning something similar?
Tell me by email what you have in mind. The first conversation is free.